Jets energy calibration in ATLAS V.Giangiobbe Università di Pisa INFN sezione di Pisa Supported by the ARTEMIS Research Training Network Workshop sui Monte Carlo, la Fisica e le Simulazioni a LHC V.Giangiobbe INFN Pisa Edition IV- February 18-20, 2008 - Laboratori Nazionali di Frascati
Jets energy calibration in ATLAS Jets are present in many signatures of interest in ATLAS Top quark physics, Higgs, susy, compositness... QCD is the background for many analysis Knowledge of the Jet Energy Scale is a crucial issue for inclusive jet cross section missing energy top quark mass (at Tevatron : most relevant systematic error on the top mass) Goal of the collaboration calibration of the Jet Energy Scale at 1% with the best possible resolution.
Outlines The calorimetry in ATLAS Strategy for jets energy calibration Reconstruction and calibration scheme Check of the jet energy scale
Calorimetry in ATLAS Electromagnetic Liquid Argon Calorimeters Tile Calorimeters Pb/LAr 3 longitudinal sections 1.2 λ η ϕ = 0.025 0.025 and higher Central Hadronic η < 1.7 Fe/scintillator 3 longitudinal sections 7.2 λ η ϕ = 0.1 0.1 and higher η=1.475 η=1.8 End Cap Hadronic 1.5 < η < 3.2 η=3.2 Hadronic Liquid Argon EndCap Calorimeters EM accordion η < 3.2 Forward Liquid Argon Calorimeters Cu/LAr 4 longitudinal sections η ϕ < 0.1 0.1 Forward calorimeter 3.1 < η < 4.9 EM Cu/LAr HAD W/Lar 3 longitudinal sections 9 λ η ϕ = 0.2 0.2
Strategy for the jets energy calibration General idea Calorimeter jet physics jet parton level Before calibration jets @ electromagnetic scale (does not mean that e and are correctly calibrated) Step 1 : corrects for instrumental effects Goal : bring reco jet energy to the total energy of MC jet particles stable (physics jet) Calorimeter and physics jets reconstructed using the same clustering algorithm corrects for instrumental effects (gap, dead material, non-compensation) JES is now independent on detector effects Step 2 : go back to initial parton energy correction for Underlying events, out-of-cluster energy... using data (in situ studies) using simulation
Strategy for the jets energy calibration Different steps of the calibration Calo cells pre-clustering Protojets Calorimeter domain Jet reconstruction domain Physics analysis domain Jet finding algorithm Uncalibrated jets Calibration Calibrated jets in-situ calibration Parton level
Reconstruction and calibration scheme From calo cells to protojets Input : calorimeter cells (uncalibrated signal) Clusters are built grouping cells in : Projective towers η ϕ = 0.1 0.1 (no longitudinal information) 3D topological clusters : groups nearest-neighbor cells around a seed with significant signal, rejects cells with insignificant signal Better noise reduction / less cells in clusters using the 3D topological clustering Output : uncalibrated protojets (towers/clusters)
Reconstruction and calibration scheme From protojets to uncalibrated jets Input : uncalibrated protojets Jet finding algorithm applied on protojets. Default algorithm for ATLAS Seeded cone algorithm iterative cone finder starting from seeds Parameters : seed threshold (typically pt=1 GeV), cone radius (in ηϕ) Kt algorithm Merge protojets which are close in space and in pt Parameter : distance D between protojets Default parameters in ATLAS : Seeded cone : R=0.4 (W and top physics), R=0.7 (QCD, jet cross-section) Kt : D=0.4 (W and top physics), D=0.6 (QCD, jet cross-section) Output : uncalibrated jets
Reconstruction and calibration scheme From uncalibrated to calibrated jets Input : uncalibrated jets Jets need to be calibrated in order to correct for dependence of the reconstructed energy (gap, dead material) non-linearity (non-compensation, inter-calibration between calorimeters) effect of the magnetic field on soft charged particles Calibration based on simulation : reference = truth jet energy E(truth) Truth jets obtained running the same jet finding algorithm as reco jets on final state particles from the event generator (except muons, neutrinos and non interacting particles) Each reconstructed jet is associated to the nearest truth jet
Reconstruction and calibration scheme From uncalibrated to calibrated jets Default in ATLAS : H1-style method reco E Calibrated jet energy calculated as : jet = i cells jet w i i E i Ei = energy deposited in i-th cell of the calorimeters ρi = energy density in cell i (energy over cell volume) wi = weighting function depending on ρi Weights obtained minimizing the quantity : Weights are computed on QCD jets sample = jets 2 reco jet E E E truth jet truth 2 jet Output : calibrated jets : JES do not depend any more (in principle) on the calorimeter characteristics
Reconstruction and calibration scheme Performance of the calibration on QCD jets Non-linearity <2% between 50 GeV and 2 TeV Energy resolution : σ E 0.64 4.9 = 0.026 E E GeV E GeV NB : Estimation based on the simulation of the detector
Alternative scheme for calibration Local hadron calibration Use calorimeter information (energy density) to distinguish between e.m. clusters and more hadronic clusters Apply a different weight depending on the type of cluster Apply the jet finding algorithm on the weighted clusters Provides better calibrated input to jet finder algorithms Still under development...
Validation of the calibration scheme Simulation of the calorimeters Jet calibration procedure relies on simulation of calorimeters response to jets. Confrontation between MC and data is a crucial step to validate the calibration scheme Many standalone and combined test-beams with calorimeters hadronic and electromagnetic end-caps Forward electromagnetic and hadronic central barrels Different types of particles :, e, p, Large range in energy (from 1 to 350 GeV) and pseudo-rapidity
Validation of the calibration scheme Simulation of the calorimeters Systematic comparison between TB data and MC : how well can simulation reproduce : total energy distribution, linearity, energy resolution shower profile (longitudinal and lateral), e/... Comparison between different versions of GEANT4 Comparison between physics lists (QGSP, + Bertini cascade, + quasi-elastic scattering...) depending the energy range Interact and iterate with physics simulation experts until required precision is achieved Define one or more "optimized and recommended" physics lists and parameters (e.g. range cuts, etc...)
Validation of the calibration scheme Simulation of the fragmentation Calibration designed to bring the JES to the total energy of particles composing the jets. Going back to the parton level requests good description of fragmentation good description of the underlying event activity A powerful tool : use the conservation of transverse impulsion in events with back-to-back objects : Pt balance Provides many informations extracted from data that can be compared with simulation effect of the jet finding algorithm jet calibration underlaying event Several strategies under study Z/ + jet : relate the hadronic scale of the jets with the well calibrated em objects QCD back-to-back di-jets (higher statistics)
Validation of the calibration scheme Simulation of the fragmentation Pt balance in + jet system Select the leading and the leading jet in the opposite hemisphere Analysis can be performed at various levels of the calibration procedure : Before the jets calibration to disentangle physics effects from calibration effects validate the fragmentation model study the underlying events After the jets calibration study of the out-of-cluster energy, underlying event jets radial profile comparative studies on jet finding algorithm Should also be possible to extract correction factor on data to bring the jet energy to the initial parton energy
Validation of the calibration scheme Simulation of the fragmentation QCD di-jet balance Select events with 2 jets back-to-back (higher statistics that Z/ + jet) Can be used to check of the uniformity vs pseudo rapidity First jet in a reference region (central region : smaller amount of dead material) Second jet as a probe for eta regions where detector geometry is more complex (gap, cracks)
Validation of the calibration scheme An interesting test on TB data Attempt to transfer MC calibration to real data Calibration factors (H1 weights) computed with MC, for pions in the TB configuration Application of these weights to real TB data after calibration read data ~3% off wrt simulation
Summary In ATLAS, jets calibration performed in two steps : Correction of instrumental effects (calibration to the jets particles level) Correction for detector independent effects (out-of-cluster energy, UE,...) Both steps are based on simulation : a good MC description is crucial detector response : use of the TB results physics : use on in-situ calibration (eg Pt balance) The validation of the calorimeters simulation is in progress MC/data comparison for electrons and pions on a large energy range comparisons between different physics lists and versions of G4 Pt balance : a tool to extract informations from data This will be used to check and validate the second step of the calibration procedure More sophisticated calibration schemes based on local hadron calibration are under development/validation.
Backup slides
Slide from Iacopo Vivarelli
Cone algorithm : slide from Michiru Kaneda Cone Algorithm (Seeded Algorithm) ET Seed::2GeV Collect neighbors around a seed in R= ( η2 + φ2 ) R=0.7::To avoid fragmentation loss for low Pt jets R=0.4::Necessary at high luminosity and to separate overlapping jets Split and Merge Merge two jets if overlapping energy is more than 50% of the least energetic jet energy. More than 50% of Ejet1 jet1 split jet2 jet1 jet2 merge
Kt algorithm
Some results from the Physics Validation Project Pions and protons at 90 in TileCal T.Carli, M.Simonyan Physics Validation Meeting Oct 17, 2007 G49 Erec/Ebeam +-2% +-2%
Some results from the Physics Validation Project ATLAS Hadronic End-cap calorimeter A.Kiryunin, P.Strizenec Physics Validation Meeting Oct 17, 2007
Some results from the Physics Validation Project Combined central calorimetry very low energy pions T.Carli Physics Validation Meeting Oct 17, 2007
More about physics lists Slide from Tancredi Carli 4 Nuclear deexcitation Evaporation etc. 3 Bertini nucleon-nucleon cascade step-like concentric nuclear potential in 3d Projectile transported along straight-lines Interaction according to free mean path 2 1 Particle nucleus collision Cross-section and angles from experiment QGS: Quark-Gluon String Nucleon is split in quark di-quark Strings are formed String hadronisation (adding qqbar pair) fragmentation of damaged nucleus with precompound (P) Nucleon/nucleon interaction+ Nuclear deexcitation Fritiof: Alternative string frag. Only momentum exchanged Parameterized models (as in old Gheisha)
Transfer of MC weights to TB data Slide from Paolo Francavilla
Pt balance with gamma+jet
Pt balance with QCD jets Example of studies realized by V.Lendermann, P.Weber, P.Hodgson